PneumoVision, an AI-based tool, was developed at IE University for robust and accurate pneumonia detection from chest X-ray images. We aim to reduce human errors in diagnosis and aid radiologists and doctors to deliver timely and precise healthcare services.
Statistics highlight pneumonia as the cause of over 3.2 million deaths annually with a significant portion attributed to misdiagnosis. We are addressing this problem by capitalizing on AI to minimize human errors and provide vision-aided diagnosis.
Our AI tool, PneumoVision, is powered by Google's Vision Transformer (ViT), a convolution-free model that focuses on important image regions using a self-attention mechanism. The model uses a transformer encoder to process split image patches.
The development of PneumoVision can aid early disease detection, enhance accuracy, and improve healthcare accessibility, thereby enabling telemedicine and remote consultations. Our AI tool can mitigate diagnostic inconsistencies and potentially be more accessible and reliable than traditional human-diagnosed methods. With our AI tool, we achieved 97.7% Accuracy on pneumonia detection.
.svg)
.png)
.png)
.png)
.png)
.png)
.png)